OpenAI reversed course on ChatGPT Pro this week. Instead of cutting the Pro $200 allowance today, it doubled usage limits across Pro and Business Premium.

ChatGPT Pro and Business Premium limits are doubled as of Oct 29. Tibo from OpenAI's Codex team posted that OpenAI is "doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today." His post covers Work and Codex usage. It gives no end date, and OpenAI hasn't published the new numbers, so all we can say is 2x the previous limits. Dots keep their extended limits "for now," and dot usage stays separate from your default usage. source

The Pro $200 cut planned for today is cancelled. Subscribers who joined between Sep 22 and Sep 29 were due to drop from 20x to 10x the Plus allowance on Oct 30. Tibo's post says OpenAI "won't be reducing usage limits for Pro 200." A follow-up from @ajambrosino said "usage limits are back to being double." If you were planning to downgrade or switch because of the cut, you don't need to. source

Codex added GPT-6.1 Terra on Oct 27. Before, GPT-5.6 Terra was the Terra option. Plus and Standard Business users now get an estimated 35 to 285 local messages per 5 hours on GPT-6.1 Terra. Credit pricing is 25 credits per 1M input tokens and 125 per 1M output tokens. If you're on Plus, try it in Codex and see how far your 5-hour window goes. source

Claude Code limits didn't change, but heavy sessions burn them faster. On Oct 28, Boris Cherny, who leads Claude Code, said Anthropic found no single root cause and no regression in Claude Code behind users hitting limits faster. He said most cases traced back to one or two heavy sessions, naming xhigh effort, /goal, Workflows, Agent Teams and long SDK jobs. Anthropic is rolling out fixes for a few smaller issues that made some sessions burn tokens faster than they should have. If your limit drains fast, check for those session types first. source

Grok subscribers can earn usage resets by referring friends. xAI's referral terms, last updated Oct 27, say "The Reward is one usage reset, which restores your included usage for the current usage period." You earn one when a friend you refer makes a successful first payment for X Premium or a higher plan. To refer, you must live in the US and have a paid Premium, Super or Ultra plan bought on the web with a credit card for at least 7 days. X Subscription Plus plans, team seats and plans billed through Apple or Google can't refer. Each reset expires 6 months after xAI adds it. source

Also this week

Mistral Large 4 is out with open weights. Mistral released the weights on Oct 29 and pitches the model at agentic coding. API pricing is $0.70 per 1M input tokens and $2.10 per 1M output tokens. source

Agent Skills now lives at the Linux Foundation. On Oct 22, Anthropic gave the Agent Skills standard to the Agentic AI Foundation, the Linux Foundation group it already gave MCP to. You can keep using or contributing to the standard without joining. source

T3 Code is adding multi-machine orchestration. An Oct 28 post says one thread can now hand work to agents on other machines, and that it's working internally so far. A follow-up says you can now control your whole fleet of machines through a single MCP server. source

Coming up

GitHub Copilot AI Credits reset on Nov 1 (UTC), so Pro, Pro+ and Max users get a fresh month of credits. source

Every change, with sources: https://codingplanlimits.com/changelog/